基于Renyi熵滤波的光声图像重建算法设计与实现

针对光声图像重建过程中存在的原始光声信号信噪比差、重建图像对比度低、分辨率不足等问题,提出了基于Renyi熵的光声图像重建滤波算法.该算法首先根据原始光声信号的Renyi熵分布情况,确定分割阈值,并滤除杂波信号;再利用滤波后的光声数据进行延时叠加光声图像重建.利用该滤波算法分别处理铅笔芯横截面(零维)、头发丝(一维)以及小鼠大脑皮层血管(二维)等不同维度样本的光声信号,实验结果表明:相比Renyi熵处理之前,重建图像对比度平均增强了32.45%,分辨率平均提高了30.78%,信噪比提高了47.66%,均方误差降低了35.01%;相比典型的滤波处理算法(模极大值法和阈值去噪法),本研究中图像的对比度、分辨率和信噪比分别提高了25.94%/10.60%、27.90%/19.48%、35.21%/10.60%,均方误差减小了28.57%/16.66%.因此,选择利用Renyi熵滤波算法处理光声信号,从而使光声图像重建质量得到大幅改善.

Design and Implementation of Photoacoustic Image Reconstruction Algorithm Based on Renyi Entropy Filter

In order to improve the quality of photoacoustic image reconstruction, aiming at the problems that the signal-to-noise ratio of the original photoacoustic signal is poor, the reconstructed image contrast is low and the resolution is insufficient in the process of photoacoustic image reconstruction, based on the quality of photoacoustic signal which collected from the optimized photoacoustic imaging system, a reconstructed filtering algorithm in view of Renyi entropy is proposed(before using the delay superposition algorithm to reconstruct the image, the original photoacoustic signal is filtered by Renyi entropy filter). Compared with the existing classical filtering algorithm (modulus maxima method and threshold denoising method), the contrast ratio of the algorithm is improved by 18.27% on average, the resolution is increased by 23.69% on average, the SNR is increased by 2.90% on average, and the mean square error is reduced by 2.61% on average by using Renyi algorithm. The photoacoustic signal of the pencil (zero-dimension), hair (one dimension) and mouse cortical blood vessels (two-dimensional) was filtered by the Renyi entropy filter before performing the photoacoustic image reconstruction. After Renyi entropy filtering, the contrast of the photoacoustic reconstructed images was greatly improved by 36.75% (pencil cross section, zero dimension), 30.22% (hairline, one dimension) and 30.38% (mouse cortical blood vessels, two-dimensional). The resolution of reconstructed images also increased significantly, but the resolution of mouse cortical blood vessels was limited (17.65%) compared with zero-dimensional and one-dimensional samples. We speculate that this is related to the selection of biological samples (the first two samples were imitation, the samples of mouse cortical blood vessels were in vivo, and the differences in the photoacoustic signals between the mouse cortical blood vessels and the surrounding biological tissues were weaker than those). The signal-to-noise ratio of reconstructed images was significantly increased by 43.20% (pencil cross section, zero dimension), 51.60% (hairline, one dimension) and 48.20% (mouse cortical blood vessels, two dimensions). Finally, the mean square error of reconstructed images decreased by 7.10% (pencil core cross section, zero dimension), 28.57% (hairline, one dimension) and 69.38% (mouse cortical blood vessels, two dimensions), with the increase of the sample dimension, the mean square error of the image is greatly reduced. We assume that this is due to the increase in the size of the sample with the increase of the sample, and the average error of the whole image is reduced, so that the mean square error corresponding to the reconstructed image is reduced. The experimental results show that the reconstructed images of the photoacoustic reconstructed by Renyi entropy compared with the reconstructed images obtained from the original photoacoustic signal, the contrast ratio of the photoacoustic reconstructed image is enhanced by 32.45%, the resolution is increased by 30.78% and the signal-to-noise ratio is increased by 47.66%, and the mean square error is reduced by 35.01%. The Renyi entropy filter processing algorithm improves the quality of photoacoustic image reconstruction which will help to promote the clinical application of photoacoustic imaging in biomedical diagnosis and treatment, for example, the early diagnosis about the arthritis, breast cancer and epilepsy and other lesions.